Raised By Robots

In the 1990s, Dazza Greenwood helped shape the laws that made electronic signatures and online contracts legally valid. Those laws already recognized something called an “electronic agent.” Thirty years later, AI agents can negotiate, make decisions and spend money on our behalf. So when your agent says yes, have you legally agreed to something?

Nolan sits down with Dazza, a lawyer and researcher at MIT and Stanford, to follow one hoodie purchase from start to finish. They cover who gave permission, whose side the agent is on, and who answers when something goes wrong.

In this episode:

• Why the e-signature laws of the ’90s anticipated today’s AI agents
• Amazon vs. Perplexity: why a store needs to know it’s really your agent, not just your password
• Interlateral: 45 lawyers working alongside their AI agents in one room at Stanford Law School
• The $49 hoodie: a bad choice, or beyond your permission?
• Accountable Agents: Dazza’s forthcoming proposal for verifying agent identity and authorization
• Loyal agents: how to test whether your shopping agent is quietly being paid to push one brand
• Cancelling an order another agent is already handling, or getting charged $80 on a $50 limit: when are you stuck paying?
• Why your identity and judgment should never be handed over to an agent

Dazza also explains why the answer to “who’s responsible?” isn’t always the person using the agent. Sometimes it’s the store, or the company providing the AI, depending on whose agent it really is.

Find Dazza’s work at dazzagreenwood.com.

What if your AI agent could book a trip, shop or spend money for you while you’re asleep? Who decides what it’s allowed to do, and who checks whether it’s making the right call?

Nolan sits down with Dick Hardt, editor of OAuth 2.0, the standard behind “Sign in with Google” and every app that asks for access to your calendar. Dick is now developing AAuth, a proposed protocol that gives each AI agent its own identity and controls what it can do on your behalf. His big idea: the agent doing the work may need a separate supervisor agent making sure it follows your instructions.

In this episode:

• Why, after 20 years, there’s still no simple way to prove who you are online
• The travel-agent problem: letting an agent grab the sushi reservation while you sleep
• Why agents should be treated as “somewhat adversarial,” and what a supervisor agent would do
• Why every copy of an agent needs its own identity, not just “I’m Claude”
• What a store actually sees: not much about the agent, but proof that a specific person authorized it
• How limits carry over when your agent brings in another agent, and how to cancel everything in milliseconds
• A starter project for a 13-year-old: build a tiny store that sells a hoodie and watch the permissions work

Dick also pushes back on Nolan’s $50 hoodie. A $55 hoodie might be the better buy, and a $20 one might be just as good. A strict budget is easy to enforce, but what you really want is good value, and that takes judgment. The challenge is giving agents enough freedom to decide well without losing control.

Find Dick’s work at aauth.dev and hello.coop.

Hugging Face is where millions of people build, share and use AI models in the open. This summer, AI agents being tested by OpenAI broke into its systems. A few weeks later, Nvidia announced a deal to acquire Hugging Face. What does that mean for the future of AI built in the open, and for the agents that are coming?

Nolan sits down with Thomas Wolf, co-founder and Chief Science Officer of Hugging Face. From physics to patent law to AI, Thomas has spent the past decade helping build one of the world's largest open-source AI communities. Last season, Nolan interviewed his co-founder, Clem Delangue.

What they cover:
• Why Hugging Face agreed to join Nvidia, and whether it can stay open
• How Hugging Face makes money while giving so much away for free
• What China's AI developers are doing differently
• Why smaller and larger AI models may work better together
• How AI could advance scientific discovery and engineering
• Why agents working together tend to agree with each other too easily
• What really happened when OpenAI's agents broke into Hugging Face
• What safety tests should prove before we trust autonomous agents
• What's still missing before robots become useful at home
• What Thomas would build if he were thirteen today

Thomas also reveals a surprising detail about the Hugging Face incident: the agents had already solved their original test, but went on a "side quest" to make sure their work would be counted. The question is no longer just whether agents can complete a task, but how far they might go to accomplish it.

Raised by Robots, Season 2: The Internet of Agents. Hosted by Nolan Cohen, 13.

If an AI agent acts for you, how does anyone know it's really yours, and how do you know it's actually working in your best interest?

Nolan sits down with Drummond Reed, who has spent 30 years working on digital identity and trust. Drummond co-edited the W3C standard for decentralized identifiers (DIDs), co-wrote Self-Sovereign Identity, helped found the Trust Over IP Foundation, and now works on the First Person Project, which is exploring how to prove there is a real person behind a digital identity or agent without relying on a centralized biometric database.

Drummond also turns the tables at the start and asks Nolan why he began interviewing AI leaders at 13.

What they cover:
• Why the internet was built without a way to know who is on the other end
• What a decentralized identifier is, and why controlling your own identifier matters
• Proving you are a real person without a centralized biometric database
• How a store could verify that an agent is really acting for you
• Fiduciary agents, and why authorization alone does not mean an agent serves your interests
• Permission slips for agents, including spending limits you can sign and revoke
• Building a trust layer for the internet
• Recovering your digital identity if you lose a device
• Portable memory, and taking what an AI knows about you to another service
• How parents and kids might share control of an agent as kids grow up
• What people Nolan's age should insist on before trusting an agent

Drummond also explains why he believes your AI memory should belong to you, not the company providing the AI, so you can inspect it, correct it and take it somewhere else.

Learn more about Drummond's work through the First Person Project and GLIA Net Alliance.

Raised by Robots, Season 2: The Internet of Agents. Hosted by Nolan Cohen.

Today's internet was built for people. We log in, search, click and buy. What changes when AI agents start doing those things for us, across different companies and platforms?

Nolan sits down with Gaowei Chang, creator of the Agent Network Protocol (ANP) and co-chair of the W3C AI Agent Protocol Community Group. After eight years at Alibaba, including work on Taobao Live, he left in 2024 to build an open protocol that lets AI agents discover each other, verify identity and communicate across platforms. He wants ANP to do for agents what HTTP did for websites.

Gaowei answered in Chinese, with live translation into English.

What they cover:
- What an Internet of Agents could look like in everyday life
- Why Gaowei left Alibaba to build an open agent network
- What today's internet is missing for agents to work across companies
- How ANP fits alongside MCP and A2A
- How an agent discovers a store it has never dealt with before
- Why agent identity and agent trust are two different problems
- How permissions and spending limits could be verified cryptographically
- What happens when your agent serves the company that built it instead of you
- Why Gaowei believes open-source personal agents will matter
- Whether large platforms will keep agents closed inside their own ecosystems
- What a 13-year-old could build now for the Internet of Agents

Gaowei also explains why he believes marketplaces like Amazon and Taobao could become less central if personal agents can discover and buy directly from stores.

Find ANP on GitHub by searching "Agent Network Protocol".
Raised by Robots, Season 2: The Internet of Agents. Hosted by Nolan Cohen.

If an AI agent shows up to buy something for you, how does the store know it is really acting for you, what it is allowed to do, and whether you actually gave it permission to spend your money?

Nolan sits down with Tim Williams, co-founder of AstraSync. Tim spent 20 years running customer operations at banks, insurers and a trading platform, working with the checks that confirm who a customer really is. Now he's building similar trust infrastructure for AI agents, through what AstraSync calls Know Your Agent.

In this episode, they look at what it would actually take for businesses to trust AI agents enough to let them buy and sell.

What they cover:
- Why human identity checks don't simply scale to AI agents
- A real-world test with agents trading on a financial exchange
- What a store can verify before letting an agent spend money
- How permissions should work when one agent hands a task to another
- What goes into an agent trust score
- Why trying to game that score can backfire
- How to trace what happened when an agent makes the wrong purchase
- How to prove who's behind an agent without exposing their private information
- Why identity alone isn't enough to make agentic commerce work
- What a 13-year-old should check before letting an agent act or spend for him

Tim also explains why the real challenge isn't just identifying an agent, but proving what it's allowed to do and how it behaves at the moment it acts.

Find Tim and AstraSync at astrasync.ai.
Raised by Robots, Season 2: The Internet of Agents. Hosted by Nolan Cohen.

AI agents are starting to act for us. But who’s watching the agents, and the companies building them?

Nolan sits down with Casey Newton, who writes Platformer, the newsletter he founded, and co-hosts the new NPR show Machine Gods with Kevin Roose. Casey has spent years reporting on the decisions made by the world’s biggest tech companies, and on what happens when those decisions go wrong.

In the first interview of Season 2, they look at what agents can actually do today, what happens when they start shopping and acting for us, and who should be responsible when things go wrong.

What they cover:

• What agents can reliably do today
• Whether a shopping agent’s recommendations can be trusted
• When stores should be allowed to block agents
• Identity, permissions and privacy when agents shop for us
• Whether every agent should be tied to a real person
• Who’s responsible when an agent buys something you never meant to buy
• What happens when agents go rogue
• Who should independently check AI companies’ claims
• Whether agents reduce or deepen Big Tech lock-in
• What AI companies should learn from social media’s mistakes with kids
• How to challenge an answer that sounds impressive but says very little

Casey also turns the questions back on Nolan, including how he thinks about AI guardrails, judgment, and where responsibility should sit between the companies building these tools and the people using them.

Find Casey at platformer.news, and Machine Gods wherever you get your podcasts.
Raised by Robots, Season 2: The Internet of Agents. Hosted by Nolan Cohen.

If your AI agent buys something from another agent, who gave it permission? How does the other side know it’s really acting for you? And who’s responsible if something goes wrong?

In this short opener to season two, 13-year-old host Nolan Cohen introduces The Internet of Agents, exploring what happens when AI moves from answering our questions to acting on our behalf.

Inspired by his season one conversation with David Klingbeil, Nolan sets out to understand how agents will communicate, earn trust, and make purchases. This season, he’ll hear from builders, researchers, and skeptics about the technology and rules taking shape, and what they could mean for the people these agents represent.

The agents are already talking. Let’s listen in.

All episodes: https://raisedbyrobots.fm

What happens when the machines stop talking to us — and start talking to each other?

Nolan sits down with David Klingbeil, former Google executive, NYU professor of AI for Marketing and Strategy, and founder of Submarine.ai, where AI agents analyze billions of video views to understand brands, trends, and culture.

David explains what actually makes an AI agent, where agents are already working today, and what comes next as they begin recommending, negotiating, and eventually buying from other agents — without a human in the loop.

They covered:

• What makes an AI agent
• What AI still needs humans for: taste and judgment
• How Submarine’s Brand Radar works
• Where agents are actually being used today
• What happens when agents talk and transact with each other
• Why standards like MCP matter
• How brands market when the customer becomes an AI
• Why human-made could become the new luxury

And David’s advice to Nolan at 13: develop taste, master AI, and build judgment.

Because maybe the question isn’t whether a robot takes your job. It’s whether you’re still in the room when the decision gets made.

What happens when stories stop being something we watch — and start becoming something we talk to?

Nolan's guest today is Ben Relles, founder of Make Believe. Before launching Make Believe, Ben helped shape internet culture through viral video, co-founding Vsauce, and leading innovation at YouTube. Now he’s exploring a new idea: entertainment that responds to you.

They talked about entertainment formats that only AI makes possible, and how his team built Reid AI — a digital version of LinkedIn co-founder Reid Hoffman.

Then they tried something that has never been done before on Raised by Robots: they brought Reid AI onto the show and interviewed it live. That led into a bigger conversation about whether personalized stories strengthen or weaken shared culture, what recreating a person with AI reveals about being human, and whether one day most of us will want AI versions of ourselves. 

They covered:

• Entertainment formats impossible before AI
• How Reid AI was built
• Interviewing an AI guest live
• Can you tell human from AI?
• Personalized stories vs shared culture
• What AI reveals about being human
• Will we want AI versions of ourselves?

Because maybe the future of entertainment isn’t about watching stories. Maybe it’s about talking to them.